“Know Your Agent” Is Becoming a New Identity Category, Says Cameron D’Ambrosi

“Know Your Agent” is developing into a distinct category within digital identity as businesses deploy AI systems that can act autonomously on behalf of people and organizations, according to FaceTec executive and longtime identity industry analyst Cameron D’Ambrosi.

Speaking on the latest episode of the ID Talk podcast, D’Ambrosi said agentic AI is raising questions about identity, authorization, and accountability that the industry does not yet have the vocabulary or infrastructure to address.

D’Ambrosi is FaceTec’s Head of Strategic Partnerships for North America and the European Union. He previously helped found Liminal and hosted more than 350 episodes of the State of Identity podcast. His earlier work included financial regulation at the New York Stock Exchange, financial services compliance, and KYC and AML consulting for major banks.

The challenge goes beyond determining whether an AI agent is legitimate. Organizations also need to establish who owns it, who authorized it, whose interests it represents, and who is responsible for its actions.

“There’s so many different layers to it,” D’Ambrosi said. “And this is where the vernacular, I think, is still lacking. What is this lingua franca of how we’re going to talk about these issues?”

In a simple case, an individual might authorize an agent to purchase a product. The relevant questions include whether it can spend money, whose money it is using, and where the goods should be delivered. Enterprise deployments could involve much more complicated chains of delegation. D’Ambrosi gave the example of one company creating an agent, a consultancy licensing it, and an end customer hiring the consultancy to use it.

“Whose agent is it in that example?” he asked. “We have the guy who owns the technology, we have the guy who rented the technology, and then we have the guy or gal who hired the company to rent the technology on their behalf.”

Such arrangements complicate how permissions are granted, who can authorize a transaction, and who bears responsibility when an agent makes a mistake or exceeds its mandate.

“You start peeling the onion, and you start blowing through layers and layers of identity questions extremely, extremely quickly,” D’Ambrosi said.

He added that enterprise users and consumers are poorly equipped to describe or manage these relationships.

“I don’t think we as an industry really have prepared both enterprise users or end consumers to even have the words to kind of discuss what is going on,” he said.

D’Ambrosi argued that AI agents make established identity and access management principles more urgent. Agents should begin with no permissions and receive access only after a specific request has been evaluated and approved.

“Every identity, whether it’s human or whether it’s agentic, starts off with, and exists in, a state of zero permission,” he said.

Access to money, code, data, and other resources should be granted on a just-in-time basis rather than through permanent credentials or standing permissions.

“Make sure there is no such thing as a standing permission that can be accessed by these agents,” D’Ambrosi said. “Because, as we can see, once they start cooking and a few wires get crossed, you have no idea what output can come out, or what they’re going to try and do or delete or spend.”

Identity has been described as the new security perimeter for years. D’Ambrosi said agentic AI is making that principle an operational requirement.

“All of the reasons why those things were so critical for human identities, agentic is just the gasoline on that fire,” he said.

D’Ambrosi also expects fraud and financial losses to shape the development of Know Your Agent frameworks. Early agentic commerce deployments will expose vulnerabilities, much as the growth of online card payments pushed merchants and payment providers to improve fraud detection and chargeback management.

“We’re going to need volume with agentic commerce transactions to find out how are fraudsters taking advantage of this, how are first-party bad actors taking advantage of this,” he said.

Those incidents will help companies identify weaknesses in their identity, authorization, and accountability systems. At the same time, D’Ambrosi cautioned against treating agentic AI as a clean break from the identity industry’s past work.

“We’re still solving the same problems, we’re still asking the same questions, and I think in many ways we’re still using the same tools to answer them and to solve for them,” he said.

The core questions remain familiar: Who is behind a transaction? Can that identity be trusted? Should the requested action be allowed? The immediate actor, however, may now be an autonomous system operating through a complex chain of delegation. Know Your Agent will therefore need to cover agent ownership, authorization, permissions, audit trails, and accountability.

Listen to the full episode through the media player below, or wherever you get your podcasts.

–

By the ID Tech Editorial Team